Submarine water intake structure hidden danger detection method based on acousto-optic joint detection

Abstract

The invention relates to the field of submarine structure detection, in particular to a submarine water intake structure hidden danger detection method based on acousto-optic joint detection. According to the method, through a multi-source data acquisition and intelligent fusion technology, including centimeter-level high-resolution sonar sounding, water body layer structure acoustic imaging, ROV optical camera shooting observation and laser calibration, scouring, crack and blockage information of the seabed grating is comprehensively obtained, and the seabed grating is comprehensively identified by adopting adaptive enhancement, threshold segmentation and spatial interpolation algorithms. The problems that a traditional sonar imaging blind area exists, the target boundary is fuzzy and the biological adhesion thickness is difficult to quantify are solved, and high-precision and multi-dimensional detection of seabed structure damage, blockage and scour forms is achieved. According to the overall scheme, the detection integrity and accuracy are improved, the submarine environment safety monitoring capacity is enhanced, and reliable technical support is provided for operation and maintenance of submarine facilities.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater structure detection, and particularly to a method for detecting potential hazards of underwater water intake structures based on combined acoustic and optical detection. Background Art

[0002] Currently, the detection of underwater structures mainly relies on acoustic imaging technology, especially devices such as side-scan sonar and multi-beam sonar. By analyzing the echo signals, an underwater structure image is constructed. Such methods have a certain penetration ability and imaging range and have been widely used in tasks such as underwater structure identification and pipeline inspection. However, traditional sonar imaging has obvious limitations. Due to the complex and variable underwater environment, the structure of the underwater structure often has features such as occlusion, mutation, or grooves, making it impossible for sound waves to effectively cover the target area, resulting in a large number of blind spots, missing image information, and incomplete structure identification.

[0003] In addition, the resolution of sonar images is limited, the ability to express details is poor, the edges of the targets are blurred, and it is difficult to meet the requirements of high-precision identification. At the same time, acoustic signals are sensitive to the environment and are easily affected by changes in water temperature, salinity, and background noise, resulting in poor image stability and low signal-to-noise ratio, further affecting the accuracy of structure positioning and size measurement. More importantly, in the existing technology, traditional detection methods are difficult to synchronously evaluate the scouring volume and blockage degree due to the single data dimension: although multi-beam sounding can identify macroscopic terrain changes, it cannot penetrate the gaps between the gratings to detect internal siltation; although sonar images can locate structural damage, the local resolution is insufficient due to the influence of sediment occlusion; while ROV optical detection is prone to image blurring under strong current conditions, and it is difficult to quantify the thickness of biological attachment.

[0004] In summary, the existing acoustic detection technology still has problems such as detection blind spots, low image quality, and insufficient identification accuracy when facing complex underwater structures and high-precision detection requirements. There is an urgent need for an improved method that integrates multi-source information and improves detection integrity and accuracy. Summary of the Invention

[0005] Aiming at the above-mentioned existing technology, the present invention provides a method for detecting potential hazards of underwater water intake structures based on combined acoustic and optical detection, mainly solving the technical problems existing in the above background art.

[0006] To achieve the above object, the technical solution of the embodiment of the present invention is realized as follows: A method for detecting potential hazards of a submarine water intake structure based on combined acousto-optic detection, the method comprising the following steps: Step S1: Using an acoustic multibeam sounding system, perform high-density scanning of the seabed around the water intake head through a broadband sonar array to obtain water surface layer topographic data with centimeter-level resolution, and calculate its overall scour volume based on the centimeter-level resolution water surface layer topographic data; Step S2: Using the high-frequency channel of a dual-frequency imaging sonar to identify millimeter-level cracks and fishing net entanglement on the surface of the grille of the water body layer structure, and dynamically adjust its signal intensity through an adaptive time-varying gain compensation algorithm to suppress image blurring caused by changes in water turbidity, and obtain millimeter-level resolution acoustic image data on the surface of the grille; Step S3: Conduct close-range observation of the seabed layer by carrying an ROV optical camera to obtain optical image data on the surface of the grille, and perform laser calibration on the optical image data to obtain laser calibration data; Step S4: Fuse the millimeter-level resolution acoustic image data on the surface of the grille with the optical image data on the surface of the grille, and apply adaptive histogram equalization to enhance the signal of the target area of the fused data, and then extract the blocked area on the surface of the grille through the Otsu threshold segmentation algorithm; Step S5: Fuse the optical image data on the surface of the grille with the laser calibration data, perform spatial interpolation compensation on the blind area on the backflow surface of the grille, extract the geometric constraints of the blind area boundary, and based on the spatial correlation variogram model of the acoustic data under the geometric constraints of the blind area boundary, and then combine with the Kriging interpolation algorithm to predict the elevation value of the blind area, and map the elevation value into the blind area to restore the scour morphology of the blind area on the backflow surface of the grille.

[0007] As a preferred solution of the present invention, the specific steps of step S1 are as follows: Step S101: Use the CUBE algorithm to preprocess the water surface layer topographic data to eliminate outliers caused by multipath effects or suspended matter interference; Step S102: Based on the reference water surface layer topographic data of the non-scoured area, use the least squares method to fit the seabed surface , and its mathematical expression is: where are all polynomial coefficients, represents coordinate direction, represents coordinate direction; Step S103: Use the Delaunay triangulation algorithm to convert the discrete points of the water surface layer topographic data into continuous triangular meshes, and calculate the scour pit volume of each surface element can be expressed as: where , , respectively represent the scour pit depths of the vertices of the triangle, represents the projected area of the triangular surface element; its overall scour pit volume is the sum of the volumes of all surface elements.

[0008] As a preferred embodiment of the present invention, the specific steps of dynamically adjusting the signal strength through the adaptive time-varying gain compensation algorithm in step S2 include: The acoustic wave intensity decays with the propagation distance: Wherein, represents the received intensity of the acoustic wave at a distance r, represents the attenuation coefficient of the received signal in water, represents the distance between the sonar and the target, represents the initial intensity of the transmitting sound source, represents the exponential function; According to and , the received gain is dynamically adjusted: Wherein, represents the real-time received gain at a distance , represents the reference gain, represents the intensity attenuation caused by geometric spreading, represents compensating for the energy loss caused by the water attenuation coefficient .

[0009] As a preferred embodiment of the present invention, step S3 further specifically includes quantifying the attachment thickness of the biological community on the grille surface based on the optical image data of the grille surface combined with the laser calibration data: Assume that the pixel height of the microscopic morphology image contour of the biological community on the grille surface in the vertical direction is , then the actual thickness is: Wherein, represents the conversion coefficient from pixels to actual distance; Then the laser calibration data specifically uses the standard deviation to quantify the thickness of the attached biological community: Wherein, represents the standard deviation of the thickness of all measurement points, represents the total number of all measurement points, represents the actual thickness of the th measurement point, represents the average thickness of all measurement points.

[0010] As a preferred embodiment of the present invention, the specific process of applying adaptive histogram equalization to enhance the target area signal of its fused data in step S4 includes: Step S401: Generate a feature map from the fused data, and then divide the feature map into sub-blocks. To avoid discontinuity between blocks, for each sub-block boundary pixels, CDF interpolation of the adjacent four blocks is used, and then each sub-block is independently equalized, and at the same time, the gray-level histogram of each sub-block is calculated Wherein, B is the gray level; Step S402: For each sub-block , generate the probability distribution of the normalized histogram: , and calculate the cumulative distribution function: , then map the original gray value to the equalized value to enhance the target area. As a preferred solution of the present invention, the specific process of extracting the blocked area on the grille surface by the Otsu threshold segmentation algorithm in step S4 includes: dividing the image gray level of the enhanced target area into two categories: blocked and unblocked, and finding the optimal threshold through the following formula to maximize the between-class variance : where represents the gray threshold, , represents the proportion of blocked pixels in the enhanced target area, represents the proportion of unblocked pixels in the enhanced target area, represents the average gray level of the proportion of blocked pixels, represents the average gray level of the proportion of unblocked pixels; then re-automatically segment the blocked and unblocked areas according to the threshold .

[0011] As a preferred solution of the present invention, step S5 specifically includes: step S501: fitting the spatial correlation of the acoustic data based on the spatial correlation variogram model of the acoustic data : where is the lag distance, represents the nugget effect, represents the sill value, represents the distance scale of the spatial autocorrelation effect; step S502: calculate the interpolation weight of the known points for the point to be interpolated according to the variogram model where is the Lagrange multiplier, represents the th known point and the th known point represents the distance between the th known point and the position of the point to be interpolated , represents the th known point and the th known point represents the spatial correlation variation value between the th known point and the position of the point to be interpolated , represents the number of known points in the blind area adjacent region, represents the th known point Position of the point to be interpolated and its interpolation weight; Step S503: Use Kriging interpolation to predict the elevation value of the blind area: wherein, represents the position of the point to be interpolated and the elevation value of the blind area, represents the elevation value of the known points in the adjacent area of the blind area.

[0012] Step S504: Map the predicted elevation value of the blind area at the position of the point to be interpolated in step S503 to the blind area on the backflow surface of the grille.

[0013] The beneficial effects of the present invention are as follows: This method uses a high-resolution multi-beam sonar for underwater topographic survey to accurately calculate the volume of the scour pit; uses a dual-frequency imaging sonar to identify structural cracks and entangled fishing nets to achieve millimeter-level detail detection; combines ROV optical images and performs laser calibration to obtain clear and high-quality optical images. By fusing multi-source data, adaptive enhancement, and threshold segmentation to extract the blocked area, it solves the problems of traditional sonar blind areas, low resolution, and blurred targets. At the same time, combined with the Kriging spatial interpolation algorithm, it compensates for the missing elevation in the blind area and restores the structural scour appearance. This method significantly improves the integrity, resolution, and accuracy of submarine structure detection, improves blind area coverage, effectively identifies potential hazards, and improves the safety of submarine facilities and the scientific nature of operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is the flow chart of the steps of this method; Figure 2 is the imaging sonar image of the acoustic multi-beam system; Figure 3 is the acoustic image of the damaged and accumulated objects of the intake head grille; Figure 4 is the optical image of the intake head grille; Figure 5 is the comparison chart before and after the enhancement of the imaging sonar image; Figure 6 is the close-up image of the grille taken by the ROV optical camera. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] The technical solution of the present invention will be further elaborated in detail below in conjunction with the accompanying drawings of the specification and specific embodiments. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. In the following description, the expression "some embodiments" is described, which describes a subset of all possible embodiments. However, it should be understood that "some embodiments" may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0016] In the following description, a large number of specific details are given to provide a more thorough understanding of the present invention. However, it is obvious to those skilled in the art that the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present invention, some well-known technical features in the art are not described.

[0017] It should be understood that the present invention can be implemented in different forms and should not be construed as limited to the embodiments presented herein. On the contrary, providing these embodiments will make the disclosure thorough and complete and will fully convey the scope of the present invention to those skilled in the art. And the purpose of the terms used herein is only to describe specific embodiments and is not a limitation of the present invention. When used herein, the singular forms "a", "an" and "the" are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms "comprising" and / or "including", when used in this specification, determine the presence of the described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups. When used herein, the term "and / or" includes any and all combinations of the related listed items.

[0018] It should be further noted that when an element is referred to as being "fixed to" another element, it can be directly on the other element or there may also be an intermediate element. When an element is considered to be "connected" to another element, it can be directly connected to the other element or there may be an intermediate element at the same time. The terms "vertical", "horizontal", "inner", "outer", "left", "right" and similar expressions used herein are only for the purpose of illustration and do not represent the only implementation manner.

[0019] In order to thoroughly understand the present invention, a detailed structure will be proposed in the following description to explain the technical solution proposed by the present invention. The optional embodiments of the present invention are described in detail as follows. However, in addition to these detailed descriptions, the present invention can also have other implementation manners.

[0020] Example 1 Please refer to the attached Figure 1 , this application provides a method for detecting potential hazards of a submarine water intake structure based on combined acoustic and optical detection. The method includes the following steps: Step S1: Use an acoustic multi-beam sounding system to perform high-density scanning of the seabed around the water intake head through a broadband sonar array, obtain water surface layer topographic data with centimeter-level resolution, and calculate its overall scouring volume based on the centimeter-level resolution water surface layer topographic data; Specifically, based on the multi-beam sounding system as the macroscopic topographic detection basis, the seabed around the water intake head is scanned at high density through a broadband sonar array (200 - 400 kHz) to generate a digital elevation model (DEM) with centimeter-level resolution. To address the problem of acoustic wave propagation being affected by fluctuations in water temperature and salinity, an AML SV+ sensor is used to correct the sound velocity profile in real time, and combined with GNSS / INS integrated navigation technology, the planar positioning accuracy is ensured to reach ±5 cm and the elevation error is less than ±2 cm. The data in this stage provides a reliable spatial reference for the macroscopic distribution of scouring pits. Specifically, the Reson T50 multi-beam sounding system (operating frequency 200 - 400 kHz, maximum sounding depth 300 m) is used in the experiment, and its swath width and water depth relationship is:

[0021] where the transducer opening angle θ = 150°; The planar positioning accuracy of ±5 cm and the elevation accuracy of ±2 cm are achieved through GNSS / INS integrated navigation. The lateral coverage width is set to 4 times the water depth to ensure complete coverage of the 20 m × 20 m area around the water intake head. To eliminate acoustic wave propagation errors, sound velocity profile data (AML SV+ sensor) is collected in real time and tidal level dynamic correction is performed. After abnormal points (confidence level > 95%) are removed by the CUBE filtering algorithm, a 0.1 m × 0.1 m grid digital elevation model is generated. See Figure 2 , the results show that there is a scouring depression area about 15 m long and 10 m wide on the northeast side of the water intake head, with an average depth of 4.2 m and a maximum scouring depth of 6.5 m. Step S2: Use the high-frequency channel of a dual-frequency imaging sonar to identify millimeter-scale cracks and fishing net entanglement on the surface of the grid of the water body layer structure, and dynamically adjust its signal intensity through an adaptive time-varying gain compensation algorithm to suppress image blurring caused by changes in water turbidity, and obtain millimeter-scale resolution acoustic image data of the grid surface; Specifically, to make up for the limitations of the water surface layer topographic data, see Figure 3, a dual - frequency imaging sonar (EdgeTech 4200MP) is introduced for detecting the water layer structure. The high - frequency 1.2 MHz channel focuses on the refined identification of millimeter - scale cracks on the grille surface and entanglement with fishing nets. The signal intensity is dynamically adjusted through an adaptive time - varying gain (TVG) compensation algorithm to suppress the image blurring caused by changes in water turbidity. The low - frequency 750 kHz channel penetrates the surface sediment to detect the siltation thickness inside the culvert, with a maximum penetration depth of 20 m. Through the spatio - temporal registration of acoustic image data and topographic data, the edge morphology of the scour pit and the damage position of the grille are accurately correlated.

[0022] The technical parameters of the multi - beam sounding system are shown in Table 1: Table 1 Technical parameters of the multi - beam system

[0023] Step S3: Use the ROV optical camera to conduct a close - range observation of the seabed layer, obtain the optical image data of the grille surface, and perform laser calibration on the optical image data to obtain laser calibration data; Specifically, referring to Figure 4 , aiming at the deficiency of acoustic means in micro - defect quantification, an ROV optical camera is deployed to the seabed layer for close - range observation. Under the strong - current condition with a flow velocity of 1.5 m / s, the ROV optical camera suppresses motion blur through an inertial - stabilized gimbal, captures the microscopic morphology of the biological community on the grille surface, and quantifies the attachment thickness (error ±0.2 mm) in combination with the laser calibration data. The fusion of optical and acoustic image data further solves the problem of missed detection of weakly - reflecting targets such as biofilms. Step S4: Fuse the millimeter - scale resolution acoustic image data of the grille surface with the optical image data of the grille surface, apply adaptive histogram equalization to enhance the signal of the target area in the fused data, and then extract the blocked area on the grille surface through the Otsu threshold segmentation algorithm; Step S5: Fuse the optical image data of the grille surface with the laser calibration data, perform spatial interpolation compensation on the blind area on the back - flow side of the grille, extract the geometric constraints of the blind - area boundary, based on the spatial correlation variogram model of the acoustic data under the geometric constraints of the blind - area boundary, and then combine with the Kriging interpolation algorithm to predict the elevation value of the blind - area region and map the elevation value into the blind area to restore the scour morphology of the blind area on the back - flow side of the grille.

[0024] As a preferred solution of the present invention, step S1 specifically includes the following steps: Step S101: Pre - process the water - surface layer topographic data using the CUBE algorithm to remove the outliers caused by multipath effects or suspended - matter interference; Step S102: Based on the reference water - surface layer topographic data of the non - scoured area, use the least - squares method to fit the seabed surface , and its mathematical expression is: where, are all polynomial coefficients, Indicates coordinate direction Indicates coordinate direction; Specifically, based on the reference terrain data of the non-scoured area, a quadratic surface model is constructed, and its fitting residual is controlled within ±3 cm. The elevation difference between the measured terrain and the reference plane is transformed into a continuous triangular mesh through the Delaunay triangulation algorithm. The final cumulative scouring volume is 625.8 m³, accounting for 18% of the designed volume of the water intake head. To verify the robustness of the algorithm, 1000 Monte Carlo simulations were carried out in the present invention, and Gaussian noise with a mean of 0 and a standard deviation of 5 cm was superimposed on the original data. The results show that the scouring volume error is stable at 3.2% - 4.7% (confidence level 95%). Sampling was carried out on the edge of the scouring pit on the northeast side through ROV optical photography, and the deviation between the measured depth and the calculated value is only 4.3%, indicating the reliability of this method under complex terrain. Specifically, the polynomial coefficients are determined by optimizing the regional terrain characteristics, and the fitting residual is controlled within ±3 cm. The elevation difference between the measured terrain surface and the reference plane is the spatial distribution of the scouring depth. To accurately calculate the volume of the scouring pit, the Delaunay triangulation algorithm is used to transform the discrete point cloud into a continuous triangular mesh. This algorithm generates an optimal triangular mesh under complex boundary conditions by maximizing the minimum interior angle criterion, effectively avoiding the integration error caused by long and narrow triangles. Step S103: Use the Delaunay triangulation algorithm to transform the discrete points of the water surface layer terrain data into a continuous triangular mesh, and calculate the scouring pit volume of each face element can be expressed as: where , , respectively represent the scouring pit depths of the vertices of the triangle, represents the projected area of the triangular face element; the overall scouring pit volume is the sum of the volumes of all face elements.

[0025] As a preferred solution of the present invention, the specific steps of dynamically adjusting its signal intensity through the adaptive time-varying gain compensation algorithm in step S2 include: The acoustic wave intensity decays with the propagation distance: where represents the received intensity of the acoustic wave at a distance , represents the water body attenuation coefficient of the received signal, represents the distance between the sonar and the target, represents the initial intensity of the transmitting sound source, represents the exponential function; according to and , dynamically adjust the received gain: where Indicates the received real-time gain at a distance , Indicates the reference gain Indicates the intensity attenuation caused by geometric spreading Indicates compensation for the energy loss caused by the water attenuation coefficient .

[0026] As a preferred embodiment of the present invention, step S3 further specifically includes quantifying the attachment thickness of the biological community on the grille surface according to the optical image data of the grille surface in combination with the laser calibration data: assuming that the pixel height of the microscopic morphology image contour of the biological community on the grille surface in the vertical direction is , then the actual thickness is: wherein represents the conversion coefficient from pixel to actual distance; then the laser calibration data specifically uses the standard deviation to quantify the thickness of the biological community attachment: wherein represents the standard deviation of the thickness of all measurement points represents the total number of all measurement points represents the actual thickness of the th measurement point represents the average thickness of all measurement points

[0027] This process improves the recognizability of the attachment layer edge through image preprocessing, and combines the precise size reference provided by laser calibration to achieve millimeter-level quantitative measurement of the thickness of the seabed biological community. This provides basic quantitative data for environmental monitoring, structural assessment and blind area repair

[0028] As a preferred embodiment of the present invention, the specific process of applying adaptive histogram equalization to enhance the target area signal of its fused data in step S4 includes: Step S401: Generate a feature map from the fused data, and then divide the feature map into sub-blocks. To avoid discontinuity between blocks, for each sub-block boundary pixels, CDF interpolation of the adjacent four blocks is used, and then each sub-block is independently equalized, and at the same time, the gray histogram is calculated for each sub-block where B is the gray level; Step S402: For each sub-block , normalize the histogram to generate a probability distribution: , and calculate the cumulative distribution function: , and then map the original gray value to the equalized value, thereby enhancing the target area As a preferred embodiment of the present invention, the specific process of extracting the blocked area on the grille surface by the Otsu threshold segmentation algorithm in step S4 includes: classifying the image grayscale of the enhanced target area into blocked and non-blocked categories, and finding the optimal threshold through the following formula to maximize the between-class variance : where represents the grayscale threshold, , represents the pixel proportion of the blocked area in the enhanced target area, represents the pixel proportion of the non-blocked area in the enhanced target area, represents the average grayscale of the pixel proportion of the blocked area, represents the average grayscale of the pixel proportion of the non-blocked area; then, according to the threshold , automatically re-segment the blocked and non-blocked areas again.

[0029] Specifically, referring to Figure 5 , for the 1.2 MHz high-frequency sonar image, apply adaptive histogram equalization to enhance the contrast of the target area, and then extract the suspected blocked area through Otsu threshold segmentation. Taking a 1 m × 1 m weak reflection area as an example, the enhanced image clearly shows the fracture contour of the grille, and the calculated area is 0.82 m². Specifically, the quantification of the grille blockage area needs to combine the acoustic image data and the optical image data of the grille surface. Use the 1.2 MHz high-frequency channel of the EdgeTech 4200MP sonar to obtain the acoustic image of the grille surface with millimeter-level resolution. In response to the decrease in image contrast caused by water attenuation, apply adaptive histogram equalization to enhance the signal of the target area, and then extract the suspected blocked area through the Otsu threshold segmentation algorithm, such as fishing net entanglement or sediment accumulation. Taking a 1 m × 1 m weak reflection area as an example, after the acoustic image is enhanced and segmented, the identified fracture area of the grille is 0.82 m²; at the same time, the ROV optical image shows that the fishing net coverage rate in this area reaches 73%, verifying the reliability of the acoustic area calculation.

[0030] Referring to Figure 6 , the ROV optical camera takes a close-up shot and shows that the fishing net coverage rate in this area reaches 73%, the biological attachment thickness is 8 - 12 mm, and the comprehensive calculated loss rate of the flow-through area is 32.1%. Among the 8 groups of grilles, 3 groups have a blockage degree exceeding the threshold.

[0031] As a preferred embodiment of the present invention, step S5 specifically includes: step S501: Based on the spatial correlation variogram model of the acoustic data to fit the spatial correlation of the acoustic data: where is the lag distance, represents the nugget effect, represents the sill value, The distance scale representing the spatial autocorrelation effect; Step S502: Calculate the interpolation weight of the known points for the point to be interpolated according to the variogram model: Wherein, is the Lagrange multiplier, represents the th known point and the th known point distance between, represents the th known point and the position of the point to be interpolated distance between, represents the th known point and the th known point spatial correlation variation value between, represents the th known point and the position of the point to be interpolated spatial correlation variation value between, represents the number of known points in the blind area adjacent region, represents the th known point interpolation weight for the position of the point to be interpolated; Step S503: Use Kriging interpolation to predict the elevation value of the blind area: Wherein, represents the elevation value of the blind area at the position of the point to be interpolated represents the elevation value of the known points in the blind area adjacent region.

[0032] Step S504: Map the elevation value of the blind area predicted for the position of the point to be interpolated in Step S503 to the blind area on the backflow surface of the grid. Specifically, there are detection blind areas in areas such as the backflow surface of the grid and the inside of the culvert box, resulting in the lack of scour terrain data. Therefore, the present invention uses ROV optical images and laser calibration data to perform spatial interpolation compensation on the blind area. By approaching and shooting with the ROV optical image to obtain a high-resolution image, combined with the laser calibration module to measure the three-dimensional coordinates of the feature points (accuracy ±0.2 mm), the geometric constraints of the blind area boundary are extracted; subsequently, combined with the spatial correlation variogram model of the acoustic data, based on the Kriging interpolation algorithm, the elevation value of the blind area is predicted. For example, due to acoustic wave occlusion, 30% of the data is missing in a certain area on the backflow surface of the grid. After optical constraint interpolation, the elevation estimation error is reduced from ±15 cm of traditional linear interpolation to ±9 cm, effectively restoring the scour morphology of this area. Specifically, through this method, combined with nearly 5 years of historical operation and maintenance data and regional geological characteristics (the bottom sediment is silty clay, median grain size ; critical incipient velocity 0.2 m / s; maximum velocity difference between flood and ebb tides 0.7 m / s), a total of 3 types of typical potential hazards are identified in this detection. The specific results are shown in Table 2: ; critical incipient velocity 0.2 m / s; maximum velocity difference between flood and ebb tides 0.7 m / s), a total of 3 types of typical potential hazards are identified in this detection. The specific results are shown in Table 2: Table 2 Hidden danger point situation and repair suggestions

[0033] Through the multi-source data fusion technology, the present invention systematically solves the problem of coupled damage diagnosis of scour and grille blockage of submarine water intake structures. Engineering practice shows that the three-dimensional detection system integrating multi-beam sounding, dual-frequency imaging sonar and ROV optical detection can effectively identify hidden defects that are difficult to capture by traditional single means.

[0034] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should be covered within the protection scope of the present invention. The protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A hidden danger detection method for submarine water intake structures based on combined acousto-optic detection, characterized in that The method includes the following steps: Step S1: Using an acoustic multi-beam sounding system, densely scan the seabed around the water intake head through a broadband sonar array to obtain water surface layer terrain data with centimeter-level resolution, and calculate the overall scour pit volume based on the centimeter-level resolution water surface layer terrain data; Step S2: Use the high-frequency channel of a dual-frequency imaging sonar to identify millimeter-level cracks and fishing net entanglement on the grille surface of the water body layer structure, and dynamically adjust its signal intensity through an adaptive time-varying gain compensation algorithm to suppress image blurring caused by changes in water turbidity, and obtain millimeter-level resolution acoustic image data of the grille surface; Step S3: Conduct a close-range observation of the seabed layer by carrying an ROV optical camera to obtain optical image data of the grille surface, and perform laser calibration on the optical image data to obtain laser calibration data; Step S4: Fuse the millimeter-level resolution acoustic image data of the grille surface with the optical image data of the grille surface, apply adaptive histogram equalization to enhance the target area signal of the fused data, and then extract the blocked area on the grille surface through the Otsu threshold segmentation algorithm; Step S5: Fuse the optical image data of the grille surface with the laser calibration data, perform spatial interpolation compensation on the blind area on the backflow surface of the grille, extract the geometric constraints of the blind area boundary, and based on the spatial correlation variogram model of acoustic data under the geometric constraints of the blind area boundary, and then combined with the Kriging interpolation algorithm, predict the elevation value of the blind area, and map the elevation value into the blind area to restore the scour morphology of the blind area on the backflow surface of the grille.

2. The hidden danger detection method for a submarine water intake structure based on acousto-optic joint detection according to claim 1, characterized in that The specific steps of step S1 include the following steps: Step S101: Preprocess the water surface layer terrain data using the CUBE algorithm to eliminate outliers caused by multipath effects or suspended matter interference; Step S102: Based on the reference water surface layer terrain data of the non-eroded area, use the least squares method to fit the seabed surface , and its mathematical expression is: Where are all polynomial coefficients, represents the coordinate direction, represents the coordinate direction; Step S103: Use the Delaunay triangulation algorithm to convert the discrete points of the water surface layer terrain data into continuous triangular meshes, and calculate the scour pit volume of each element can be expressed as: Where , , respectively represent the scour pit depths of the vertices of the triangle, represents the projected area of the triangular element, and the overall scour pit volume is the sum of the volumes of all elements.

3. The hidden danger detection method for a submarine water intake structure based on acousto-optic joint detection according to claim 2, characterized in that, In the step S2, the process of dynamically adjusting its signal strength through the adaptive time-varying gain compensation algorithm specifically includes: the acoustic wave intensity decays with the propagation distance: Among them, represents the received intensity of the acoustic wave at a distance r, represents the water attenuation coefficient of the received signal, represents the distance between the sonar and the target, represents the initial intensity of the transmitting sound source, represents the exponential function; according to and , the received gain is dynamically adjusted: Among them, represents the real-time received gain at a distance , represents the reference gain, represents the intensity attenuation caused by geometric diffusion, represents compensating for the energy loss caused by the water attenuation coefficient .

4. The hidden danger detection method for a submarine water intake structure based on acousto-optic joint detection according to claim 3, wherein, The step S3 specifically further includes quantifying the attachment thickness of the biological community on the grille surface based on the optical image data of the grille surface in combination with the laser calibration data: Assuming that the pixel height of the contour of the microscopic morphology image of the biological community on the grille surface in the vertical direction is , then the actual thickness is: wherein, represents the conversion coefficient from pixels to actual distance; then the laser calibration data specifically uses the standard deviation to quantify the thickness of the biological community attachment: wherein, represents the standard deviation of the thickness of all measurement points, represents the total number of all measurement points, represents the actual thickness of the th measurement point, represents the average thickness of all measurement points.

5. The hidden danger detection method of a submarine water intake structure based on acousto-optic joint detection according to claim 4, characterized in that, The specific process of applying adaptive histogram equalization to enhance the target area signal of the fused data in step S4 includes: step S401: generating a feature map from the fused data, and then dividing the feature map into To avoid discontinuity between blocks, The boundary pixels are interpolated using the CDF of the four adjacent blocks, and then for each sub-block Equalize each sub-block independently. Calculate grayscale histogram , where B is the gray level; Step S402: for each sub-block , the normalized histogram generates the probability distribution: , and calculate the cumulative distribution function: , and then map the original grayscale value to the equalized value to enhance the target area.

6. The hidden danger detection method for a submarine water intake structure based on acousto-optic joint detection according to claim 5, characterized in that The specific process of extracting the blocked area on the grille surface by the Otsu threshold segmentation algorithm in step S4 includes: dividing the image grayscale of the enhanced target area into two categories, blocked and unblocked, and finding the optimal threshold through the following formula to maximize the between-class variance : where represents the grayscale threshold, , represents the pixel proportion of the blocked area in the enhanced target area, represents the pixel proportion of the unblocked area in the enhanced target area, represents the average grayscale of the pixel proportion of the blocked area, represents the average grayscale of the pixel proportion of the unblocked area; then, according to the threshold the blocked and unblocked areas are automatically segmented again.

7. A hidden danger detection method for a submarine water intake structure based on acousto-optic joint detection according to claim 6, characterized in that, The specific steps of step S5 include: Step S501: Based on the spatial correlation variogram model of acoustic data to fit the spatial correlation of acoustic data: where is the lag distance, represents the nugget effect, represents the sill value, represents the distance scale of the spatial autocorrelation effect; Step S502: Calculate the interpolation weight of the known points for the point to be interpolated according to the variogram model: where is the Lagrange multiplier, represents the th known point and the th known point represents the th known point and the position of the point to be interpolated distance between, represents the th known point and the th known point represents the th known point and the position of the point to be interpolated spatial correlation variation value between, represents the number of known points in the blind area adjacent region, represents the th known point pair of the position of the point to be interpolated interpolation weight; Step S503: Use Kriging interpolation to predict the elevation value of the blind area: where represents the elevation value of the blind area at the position of the point to be interpolated ; represents the elevation value of the known points in the blind area adjacent region; Step S504: Map the elevation value of the blind area at the predicted position of the point to be interpolated in step S503 to the blind area on the backflow surface of the grille.

Images